langchain-agentoracle

LangChain integration for AgentOracle — trust verification tools for AI agents.

Installation

pip install langchain-agentoracle

Quick Start

from langchain_agentoracle import AgentOracleVerifyTool, AgentOracleEvaluateTool

# Free preview verification
verify = AgentOracleVerifyTool()
result = verify.invoke("GPT-5 was released in March 2026")

# Full multi-source evaluation ($0.02 USDC via x402)
evaluate = AgentOracleEvaluateTool()
result = evaluate.invoke({
    "content": "Bitcoin reached $100K in 2025. Ethereum moved to proof-of-stake in 2022."
})

Use with LangChain Agents

from langchain_openai import ChatOpenAI
from langchain.agents import AgentExecutor, create_openai_tools_agent
from langchain_agentoracle import AgentOracleVerifyTool, AgentOraclePreviewTool

llm = ChatOpenAI(model="gpt-4")
tools = [AgentOracleVerifyTool(), AgentOraclePreviewTool()]

# Agent can now verify claims before acting on them

Tools

How It Works

AgentOracle uses 4-source verification:

1. Sonar — Primary research
2. Sonar Pro — Deep research
3. Adversarial — Actively tries to disprove claims
4. Gemma 4 — Claim decomposition and confidence calibration

Returns per-claim verdicts (supported/refuted/unverifiable) with confidence scores (0.00-1.00).

Links

• API: agentoracle.co
• npm SDK: agentoracle-verify
• GitHub: TKCollective/x402-research-skill

License

MIT